2 ms·
As others have commented, the videos look great, and much closer to how people remember journeys. However, there appear to be some image persistence problems (
by bsimpson 12y ago
As others have commented, the videos look great, and much closer to how people remember journeys. However, there appear to be some image persistence problems (many street poles simply dissolve as they get closer to the camera).
I'm curious to see what happens if they insert more action-packed footage. An MTB course with trees, switchbacks, and jumps would be an interesting stress test of this technique.
- sp332 12y agoThey do frame selection first, then create a 3D scene from the fewer frames. So that light pole might not have been in enough of the chosen frames to get a good 3D model of it. And the first stage has a pretty low-res 3D model because the number of points just gets crazy, so the light pole probably wouldn't have been modeled anyway.